POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER - Platforms' New Efforts to Block Offensive Clips Explained - Meredith
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER Platforms’ New Efforts to Block Offensive Clips Explained JUNE 2018 Paul Verna Contributors: Annalise Clayton, Sean Creamer
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER:
PLATFORMS’ NEW EFFORTS TO BLOCK OFFENSIVE CLIPS EXPLAINED
YouTube, Facebook and Twitter have faced a deluge of marketer objections over offensive videos and other
problematic content on their platforms. In the past year alone, these services have found themselves at the center
of controversies over everything from live streamed suicides to videos that promote terrorism to racist tweets from
celebrities and others.
These incidents—combined with a backlash over a WHAT’S IN THIS REPORT? This report explores how
range of related issues including promoting fake news, YouTube, Facebook and Twitter are policing user-uploaded
providing a forum for abusive trolls, influencing the 2016 video to block content that violates their guidelines.
US presidential election and misusing customer data— These platforms have recently stepped up their efforts
have put the platforms on the defensive. to promote brand safety in the wake of scandals relating
to offensive content and data privacy, which in turn led to
■■ Over the past few months, each platform has taken ad pullbacks.
steps to beef up security and promote brand safety,
particularly around video. These efforts have involved a
combination of human monitors and machine learning
analytics—with an emphasis on the latter, given the
scope of the issue.
■■ The companies have also taken steps to provide more CONTENTS
transparency to advertisers, regulators, users and
anyone else wanting to know how the platforms deal 2 Policing Video Content on YouTube, Facebook and Twitter:
Platforms’ New Efforts to Block Offensive Clips Explained
with troublesome content.
4 Brand Safety Under Threat
■■ There are indications that these companies are making
4 YouTube
progress in beating back the tide of offensive content
through improved algorithms and a sharp focus on the 9 Facebook
issue, but most experts say no platform will ever be 12 Twitter
100% safe as long as it traffics in user-generated media. 14 Brand Safety: A Shared Responsibility
■■ Despite ad boycotts, public outcries, regulatory scrutiny 14 Key Takeaways for Marketers
and punishing media coverage, the brand safety issues 15 eMarketer Interviews
that have plagued these platforms have not had an
16 Read Next
apparent effect on their bottom lines or share prices so
far. Marketers continue to see value in advertising on 16 Sources
these platforms. 16 Editorial and Production Contributors
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 2
YOUTUBE’S VAST VIDEO LANDSCAPE
2,400 Amount of
video uploaded
every minute of
KEY STAT
The amount of video
MINUTES “real” time*
that passes through
2018
2017
1996
1995
1974
1973
1952
1951
1930
1929
1908
1907
1886
1885
1864
1863
1842
1841
1820
1819
1798
1797
1776
1775
1754
1753
YouTube on any given
1732
1731
day is staggering.
2016 1994 1972 1950 1928 1906 1884 1862 1840 1818 1796 1774 1752 1730
2015 1993 1971 1949 1927 1905 1883 1861 1839 1817 1795 1773 1751 1729
144,155
2014 1992 1970 1948 1926 1904 1882 1860 1838 1816 1794 1772 1750 1728
Combined running 2013
2012
1991
1990
1969
1968
1947
1946
1925
1924
1903
1902
1881
1880
1859
1858
1837
1836
1815
1814
1793
1792
1771
1770
1749
1748
1727
1726
time of videos
2011
2010
2009
2008
1989
1988
1987
1986
1967
1966
1965
1964
1945
1944
1943
1942
1923
1922
1921
1920
1901
1900
1899
1898
1879
1878
1877
1876
1857
1856
1855
1854
1835
1834
1833
1832
1813
1812
1811
1810
1791
1790
1789
1788
1769
1768
1767
1766
1747
1746
1745
1744
If a day’s clips were
1725
1724
1723
1722
watched back to
2007 1985 1963 1941 1919 1897 1875 1853 1831 1809 1787 1765 1743 1721
watched every day**
2006 1984 1962 1940 1918 1896 1874 1852 1830 1808 1786 1764 1742 1720
YEARS
2005 1983 1961 1939 1917 1895 1873 1851 1829 1807 1785 1763 1741 1719
2004 1982 1960 1938 1916 1894 1872 1850 1828 1806 1784 1762 1740 1718
2003 1981 1959 1937 1915 1893 1871 1849 1827 1805 1783 1761 1739 1717
back, it would take
2002 1980 1958 1936 1914 1892 1870 1848 1826 1804 1782 1760 1738
2001 1979 1957 1935 1913 1891 1869 1847 1825 1803 1781 1759 1737
2000 1978 1956 1934 1912 1890 1868 1846 1824 1802 1780 1758 1736
1999 1977 1955 1933 1911 1889 1867 1845 1823 1801 1779 1757 1735
150 millennia (not
1998 1976 1954 1932 1910 1888 1866 1844 1822 1800 1778 1756 1734
1997 1975 1953 1931 1909 1887 1865 1843 1821 1799 1777 1755 1733
millennials!) to go
70 million Total watch time of
videos created in through them.
HOURS YouTube Spaces**
Views of
“Despacito” 5.24
on YouTube** BILLION
Sources: * The Wall Street Journal, “YouTube Tops 1 Billion Hours of Video a Day,
on Pace to Eclipse TV,” Feb 27, 2017; ** YouTube; “Despacito” view count as of
June 19, 2018; eMarketer calculationsBRAND SAFETY UNDER THREAT The core issues that these platforms have faced are
different, but they share something in common: They
threaten the image the companies have tried to cultivate
In early 2017, YouTube found itself grappling with a
as spaces in which average citizens can speak their
flurry of issues that threatened its image as a brand- minds, and marketers can reach vast numbers of people.
safe haven for video and display advertising. In
February, PewDiePie—at the time the YouTuber with There’s another common thread throughout these brand
the most subscribers—posted anti-Semitic videos safety issues. Despite actual and threatened boycotts,
on his channel. A month later, YouTube inadvertently public outcries, testimony before the US Congress and
UK Parliament, and a general feeling of unease with how
ran ads from major companies next to videos
these platforms have managed—or even accepted—their
supporting terrorism. roles in the digital media ecosystem, their businesses
have flourished while they faced serious existential
YouTube responded quickly to these crises by severing
threats. There’s no guarantee that business success will
ties with PewDiePie, removing the offensive videos and
continue, but if the past 15 months are any indication,
apologizing to marketers. However, these measures
YouTube, Facebook and Twitter have proven, if nothing
did not stop major brands, including GlaxoSmithKline,
else, that they are resilient platforms that deliver value to
Johnson & Johnson, Enterprise, AT&T and Verizon, from
users and advertisers alike.
pulling their ads from the platform, at least temporarily.
The backlash gathered steam throughout the first quarter
of last year, and soon hundreds of companies were
pulling back from YouTube. In late March 2017, an analyst YOUTUBE
at Nomura Instinet estimated that the boycott would cost
Google up to $750 million. Addressing a capacity crowd at YouTube’s annual
Brandcast event on May 3, 2018, CEO Susan Wojcicki
Over the course of the year, YouTube continued to take
acknowledged the company’s struggles with brand
steps to prevent brand safety lapses, and the issue died
down. By October, a Credit Suisse report noted that most safety over the past year.
advertisers had returned to YouTube, and that the issues
had not caused Google to lose any business. “With openness also come challenges, as some have
tried to take advantage of our services,” she said. “There
Other brand safety issues followed, however, and again isn’t a playbook for how open platforms operate at our
they were met by boycotts or threats from advertisers. At scale. But the way I think about it is, it’s critical that we
its May 2018 Digital Content NewFronts presentation to are on the right side of history.”
advertisers, YouTube again found itself on the defensive,
as it had a year earlier. In Q2 2018, the company released It was the second year in a row Wojcicki was compelled
its first-ever Community Guidelines Enforcement report, to bring up a topic she’d likely rather avoid. In May 2017,
designed to provide transparency into its video policing after some brands had started boycotting YouTube over
efforts and reassure advertisers that it is working hard to the inadvertent placement of their ads next to offensive
avert further failures. videos, Wojcicki apologized to Brandcast attendees.
During the same time span, Facebook and Twitter also “The last couple weeks have been challenging for some
grappled with their share of issues around data security of you,” she said at the event, which is designed to
and brand safety. Starting in late 2016, both companies showcase content for advertisers. “I want you to know
were heavily scrutinized for their purported roles in that we have taken your feedback to heart. We work hard
spreading fake news and potentially influencing the US every day to earn our advertisers’ and agencies’ trust. We
presidential election. Then, in early 2018, Facebook was apologize for letting some of you down. We can, and we
rocked by the Cambridge Analytica scandal, in which the will, do better.”
UK-based data firm with ties to the Trump campaign was
The resurgence of the brand safety issue belies the steps
found to have misused data from as many as 87 million
YouTube has taken to shield its platform against content
Facebook users in the context of the 2016 election.
that violates its terms of service. Over the past year,
the company has hired as many as 10,000 staffers to
monitor content, and invested in artificial intelligence (AI)
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 4technology to identify and block potentially problematic Share of YouTube Videos Worldwide that Are Flagged
videos. It has also lived up to its pledge to be more by Humans, by Reason for Flagging, Q4 2017
transparent, launching the first of a series of reports in % of total
April 2018 that shed light on its policing efforts. Sexual 30.1%
Spam or misleading 26.4%
YouTube removed 8.3 million videos in Q4 2017, according
Hateful or abusive 15.6%
to the YouTube Community Guidelines Enforcement
report, which the company said it would issue quarterly Violent or repulsive 13.5%
starting in April. More than 80% of those videos were 7.6% Harmful or dangerous acts
flagged by the platform’s automated systems, while 5.2% Child abuse
13.7% were flagged by an “individual trusted flagger.”
1.6% Promotes terrorism
Average users accounted for just under 5% of flags,
and NGOs and government agencies made up less than Note: total human flagged videos=9.3 million; a single video may be flagged
multiple times and for different reasons; flagging a video does not
1% each. necessarily result in it being removed
Source: YouTube, "YouTube Community Guidelines Enforcement," April 23,
2018
237492 www.eMarketer.com
Number of YouTube Videos Removed Worldwide, by
Flagger Type, Q4 2017 Over three-quarters of videos removed by YouTube in Q4
Automated flagging 2017 were not viewed at all before being blocked, while
6,685,731 the remainder received some number of views.
Individual trusted flagger
1,131,962
Share of YouTube Videos Worldwide that Are
User
Removed Before vs. After Receiving Any Views,
402,335 Q4 2017
NGO % of total
63,938
Government agency
73 After
24.1%
Note: total videos removed=8.3 million
Source: YouTube, "YouTube Community Guidelines Enforcement," April 23,
2018
237490 www.eMarketer.com
Before
Of the videos blocked by humans, the largest percentage 75.9%
were flagged for sexual content, followed by spam or
misleading content. Other videos were flagged because
Note: total videos removed=8.3 million
they were considered hateful or violent, or because Source: YouTube, "YouTube Community Guidelines Enforcement," April 23,
they depicted harmful or dangerous acts, or child abuse. 2018
237491 www.eMarketer.com
Videos that promote terrorism made up only 1.6% of
flags, despite the disproportionate attention this category Although YouTube did not provide details of how this
received in some of YouTube’s most publicized brand metric may have shifted over time, there are indications
safety issues. that the company’s machine learning processes are
getting better and faster at blocking videos before they’re
seen by large numbers of users. In Q1 2018, more than
half of the clips taken down for depictions of violent
extremism were blocked before they were seen by 10
or more viewers. By comparison, in Q1 2017 the vast
majority of such videos were viewed by at least 100
people before being removed.
Industry executives are generally reacting positively to
YouTube’s policing efforts, as well as the transparency
the company has shown by publishing its Community
Guidelines Enforcement report.
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 5“YouTube has done a really good job of cleaning out US YouTube Net Video Ad Revenues, 2016-2020
the content that really shouldn’t be on the platform at billions, % of YouTube revenues and % of US video ad
all, or shouldn’t be carrying ads,” said Mike Henry, CEO spending
$3.88
of advertising analytics firm OpenSlate. “And I think
$3.57
that’s really important because the issue, as it started
74.0% 74.0% $3.23
in February of last year, was about brand safety. While
acknowledging that even their system can’t be perfect on 73.0% 72.0% 71.0%
$2.87
that front, they have done all they can to clean it up.” $2.16
Others say recent steps are positive, but that YouTube
needs to do more.
“It’s a good first step,” said Matthew Fanelli, senior 20.0% 19.7% 18.1% 16.8% 16.0%
vice president at MNI Targeted Media. “The curtain has
been opened about a foot, and what people really want 2016 2017 2018 2019 2020
is to have it open 6 feet. They want to know YouTube is YouTube net video ad revenues
excluding all the explicit stuff, but they’re also saying, ‘I % of YouTube revenues % of US video ad spending
want a purview, and I want a look into every place that Note: net ad revenues after company pays traffic acquisition costs (TAC)
and content acquisition costs (CAC) to partner sites
I’m going to be.’ That’s the goal of where marketers want Source: company reports; eMarketer, March 2018
to get to.” 236044 www.eMarketer.com
Although our estimates of YouTube revenues themselves
Even as YouTube has grappled with brand safety
haven’t changed, we made a downward revision in the
challenges, the company’s business results have
company’s share of the US video ad pie. We now expect
been healthy.
this figure to drop from 18.1% in 2018 to 16% in 2020.
eMarketer’s March 2018 forecast of US YouTube net video In our September forecast, we had projected YouTube’s
ad revenues was unchanged from the previous version in market share to remain constant in the range of 20%
September 2017. We expect the platform to reach $3.88 to 22% for the entire forecast period. The reasons for
billion in US video advertising in 2020, up from $3.23 the change are simple: Overall video ad spending has
billion in 2018. Further, we expect video’s percentage of increased faster than we expected, and the market has
overall YouTube revenues to decrease slightly in that same grown more competitive, with more players vying for
time frame, from 73% in 2018 to 71% in 2020. This is video ad dollars.
because of the prevalence of overlays and other nonvideo
formats on the service.
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 6YouTube’s viewer numbers have also continued to US Companies Using YouTube for Marketing
flourish despite the company’s brand safety skirmishes. Purposes, 2015-2019
Its worldwide audience is expected to reach 1.86 billion % of total
51.9%
in 2021, up from 1.58 billion in 2018. As these numbers 48.2%
50.2%
grow by single-digit percentages in the next three years, 45.6%
42.7%
they will also make up a slightly larger share of total video
viewers, indicating that YouTube will not only grow its
ranks around the world but also capture a larger market
share as it goes.
YouTube Viewers and Penetration Worldwide,
2016-2021
billions, % change and % of digital video viewers
1.86
1.78
1.68 2015 2016 2017 2018 2019
1.58
1.47 Note: companies with 100+ employees; includes use of any of the
1.35 proprietary public-facing tools on YouTube for marketing purposes; includes
organic and paid functions
Source: eMarketer, Nov 2017
66.3% 66.2% 66.4% 66.5% 66.6% 66.8% 231896 www.eMarketer.com
These indicators point to YouTube’s resilience in the face
of the brand safety challenges it has faced since at least
early 2017. Despite predictions from at least one analyst
13.0%
9.2% 7.5% 6.6% 5.6% 4.9%
at that time that ad boycotts could cost parent company
Google up to $750 million, the company has seemingly
2016 2017 2018 2019 2020 2021 not missed a beat.
YouTube viewers % change % of digital video viewers
Google’s Q1 2018 earnings rose to $31.15 billion, a 25.8%
Note: individuals of any age who watch video via YouTube app or website
at least once per month increase over the Q1 2017 total of $24.75 billion. Similarly,
Source: eMarketer, Jan 2018
its closing share price on May 1, 2018, was $1,060.32,
234682 www.eMarketer.com
which represented a 33.1% increase over the January 1,
YouTube also shows no sign of slowing down when it 2017, closing of $796.79. Although Google does not break
comes to companies using the platform for marketing out YouTube results, eMarketer estimates that YouTube’s
purposes, including ads and organic posts. Our November net worldwide ad revenues totaled $7.80 billion in 2017, a
2017 forecast indicated that 51.9% of US companies 39.8% jump over the 2016 figure of $5.58 billion. These
would use YouTube for marketing in 2019, up from 48.2% are net figures after subtracting traffic acquisition and
in 2017. The forecast occurred long after a wave of brand content acquisition costs to partner sites.
safety issues on YouTube in H1 2017.
YouTube Net Ad Revenues Worldwide, US vs. Non-US,
2016-2020
billions and % of total
2016 2017 2018 2019 2020
US $2.92 $3.88 $4.43 $4.96 $5.47
—% of worldwide total 52.3% 49.7% 48.5% 47.2% 46.5%
Non-US $2.66 $3.93 $4.70 $5.54 $6.29
—% of worldwide total 47.7% 50.3% 51.5% 52.8% 53.5%
Worldwide $5.58 $7.80 $9.13 $10.50 $11.76
Note: net ad revenues after company pays traffic acquisition costs (TAC)
and content acquisition costs (CAC) to partner sites; includes banners, rich
media, search and video ads
Source: company reports; eMarketer, March 2018
236043 www.eMarketer.com
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 7YOUTUBE’S BRAND SAFETY CHALLENGES HAVE NOT AFFECTED BUSINESS RESULTS
GOOGLE WORLDWIDE REVENUES
Q1 2017 – Q1 2018 +25.8%
GOOGLE SHARE PRICES
Jan 1, 2017 – May 1, 2018 +33.1%
Some marketers
suspend YouTube
Top YouTuber PewDiePie dropped campaigns after
after posting anti-Semitic videos ads run next to child YouTube
YouTube touts exploitation videos; suspends ads
“better detection YouTube takes down on Logan Paul’s
Brands ditch YouTube after ads hundreds of channels after
run next to objectionable content; and faster removal
driven by machine accounts and vlogger posts
analyst estimates boycott will cost removes 150,000 footage of a
Google $750 million learning”
videos suicide victim
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR
2017
2018
YouTube raises Credit Suisse says ad YouTube again CNN reports that
thresholds for video suspensions had no raises video YouTube ran ads
monetization under effect on Google’s monetization from hundreds of
YouTube Partner business, and that most thresholds in an brands next to
Program (YPP) advertisers returned effort to ward off extremist content
“bad actors”
YouTube debuts quarterly
Community Guidelines Enforcement
report detailing its video policing efforts
Source: Compiled by eMarketer from company reports and Business Insider, CNBC, Cnet, CNN, MarketWatch, The New York Times, Variety, The Wall Street JournalWhile the YouTube figures are a relatively small portion The first study focused on six areas of violative content:
of Google’s total revenues—which include search and graphic violence, adult nudity/sexual nudity, terrorist
display advertising, and also ancillary sources such propaganda, hate speech, spam and fake accounts. For
as devices—it’s notable that both YouTube and its each of those content types, the report attempted to
parent company posted significant revenue increases answer four questions:
during a year when they faced a battery of brand
safety challenges. ■■ How prevalent are Community Standards violations
on Facebook?
These positive results reflect YouTube’s enduring appeal
to marketers, according to Todd Krizelman, CEO and co- ■■ How much content does the platform take action on?
founder of marketing analytics firm MediaRadar. ■■ How much violating content does Facebook find before
users report it?
“The moment the unsavoriness is discovered in the
press, the agency that represents the brand, or the ■■ How quickly does the company take action
brands themselves, will immediately disavow their on violations?
relationship with any broadcaster or publisher that
exposed them to the liability,” Krizelman said. “What I By far the most prevalent types of noncompliant content
think is interesting is how ephemeral their reaction is. on Facebook are spam and fake accounts, which exceed
If someone is doing a large campaign and it turns out other types of content by orders of magnitude. In Q1
someone catches unsavory video content, they will pull 2018, Facebook took action on 836.0 million instances
it, but then they’ll come back, sometimes within a week of spam, up from 727.0 million in Q4 2017. Similarly,
or just a few days. Very few companies say, ‘You know Facebook acted on 583.0 million cases of fake accounts
what? We’re really turned off by this forever and we’re in Q1 2018, down from 694.0 million the previous quarter.
just going to stop working with you.’”
With a time span of only two quarters, and without much
To Krizelman’s point, companies that initially stepped back context around the data, it’s difficult to put it into broader
from YouTube but eventually returned to the platform perspective and see how these types of content, and
include Procter & Gamble, Ford Motor Co., PepsiCo, Hulu Facebook’s response, may have evolved over time. It is
and Groupon, according to an article in Business Insider also unclear why cases of spam increased while fake
on May 3, 2018. accounts decreased in Q1 2018. But given the sheer
numbers of bogus accounts, flagging and taking them
down is a massive challenge for Facebook. As the study
indicates, fake accounts are “often the starting point for
FACEBOOK other types of standards violations,” so keeping them in
check presumably helps stem the tide of other types of
In May 2018, Facebook published the Community problem content downstream.
Standards Enforcement Preliminary Report, the first
in what it says will be an ongoing series of reports Number of Content Pieces that Violate Facebook
detailing its efforts to flag and block offensive content Standards, by Violation Type, Q4 2017 & Q1 2018
millions
of all types: photos, videos, images and comments.
Q4 2017 Q1 2018
The report covered the period from October 2017 Fake accounts 694.0 583.0
through March 2018, and future reports—which Spam 727.0 836.0
will be released periodically but not necessarily Adult nudity and sexual activity 21.0 21.0
Graphic violence* 1.2 3.4
quarterly—will also allow readers to look back on
Hate speech* 1.6 2.5
earlier data for comparison.
Terrorist propaganda* 1.1 1.9
Note: violating content on which Facebook took action (e.g., removing a
piece of content from Facebook, covering photos or videos that may be
disturbing to some audiences with a warning, disabling accounts); includes
posts, images, videos or comments; *increase is primarily due to
improvements in detection technology
Source: Facebook, "Community Standards Enforcement Report," May 15,
2018
238384 www.eMarketer.com
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 9The amount of content depicting adult nudity and sexual While these figures don’t specifically relate to video, Facebook’s
activity was constant during the study period, with 21.0 decision to consolidate its premium video programming under
million pieces in each of the two quarters tracked. Content the Watch tab gives brand marketers reassurance that at least
featuring graphic violence, terrorist propaganda and hate some forms of video are shielded from the types of violations
speech was less prevalent, with numbers in the single-digit that can occur more easily with user-generated content,
millions, but all three of those categories saw significant including live streams via Facebook Live.
increases, with graphic violence instances nearly tripling
from 1.2 million in Q4 2017 to 3.4 million in Q1 2018. Facebook Watch launched in mid-2017 and is ramping up
this year with entertainment shows featuring the likes of
Again, the data set doesn’t give enough context to draw Jada Pinkett Smith, and news programming under the
conclusions as to why some content types fluctuate more direction of the company’s head of news partnerships,
than others. A longer look back at this data might help former TV news anchor Campbell Brown.
explain, for example, if occurrences of hate speech have
increased as online discourse has grown more polarized “The more that you can curate, then the less the chances
in the US and elsewhere, but otherwise rationalizing the of things going wrong,” said Susan Schiekofer, chief digital
ups and downs in these categories is challenging at best. investment officer at GroupM. “Facebook is leaning in
a big way into Watch. They’re working with partners.
What is clear from other data in the study, though, is that They’re renewing shows. That is a more brand-safe video
Facebook is making progress in some of the areas that environment for clients.”
saw the biggest increases in violative activity during the
study period. In Q1 of this year, the company flagged The challenges that Facebook has encountered since
85.6% of graphic violence instances before they were early 2017 include accusations of fake news, interference
reported by users, up from 71.6% the previous quarter. It in the 2016 election, instances of violent acts streamed
also increased the rate at which it flagged hate speech, to on Facebook Live and the Cambridge Analytica scandal.
38.0% from 23.6%. Despite these challenges, which affected average users
and brand marketers alike, Facebook’s business did not
suffer during this period.
Share of Violating Content that Are Flagged by
Facebook Before Reported by Users, by Violation
Type, Q4 2017 & Q1 2018 Its Q1 2018 earnings totaled $11.97 billion, a 49.0%
% of total increase from the Q1 2017 figure of $8.03 billion. Its
Q4 2017 Q1 2018 closing share price on May 1, 2018, was $185.74, up
Spam 99.8% 99.7% 42.5% over the $130.32 closing on January 1, 2017.
Terrorist propaganda* 96.9% 99.5%
Fake accounts 99.1% 98.5% In addition to posting robust earnings and value to
Adult nudity and sexual activity* 94.4% 95.8% shareholders during the period in question, Facebook also
Graphic violence* 71.6% 85.6% emerged unscathed from public campaigns such as the
Hate speech* 23.6% 38.0%
#DeleteFacebook hashtag, which arose in the immediate
Note: read as 71.6% of content that violated graphic violence standards
and were acted on during Q4 2017 was flagged by Facebook before users aftermath of the Cambridge Analytica crisis but never
reported it; violating content on which Facebook took action (e.g., gained enough momentum to make a serious dent in
removing a piece of content from Facebook, covering photos or videos that
may be disturbing to some audiences with a warning, disabling accounts); user numbers.
includes posts, images, videos or comments; *increase due to improved
technology and processes to find and flag content before users report it
Source: Facebook, "Community Standards Enforcement Report," May 15, “Amidst all this scrutiny, Facebook is still doing really
2018
238385 www.eMarketer.com
well,” said Serena Budhwani, account director at creative
agency Sandbox. “They’ve seen ad spending go up
Clearly, Facebook has more work to do to bring those significantly on the platform, and they’re confident
figures up to the 90% range, as it has done with nudity, enough to move forward with the release of [a] dating
terrorist propaganda, spam and fake accounts. But the app. It’s not like people are leaving in droves.”
improvements in its early flagging of violence and hate
speech indicate that its algorithms are getting better at
recognizing those types of content. With hate speech in For more on Facebook’s struggles with data privacy and
particular, even recognizing it, or allowing people to repost brand safety, see eMarketer’s June 2018 report, “Changes
an instance of it to make a counterargument, can present to Facebook Advertising After Cambridge Analytica: What’s
a huge challenge. Happening with Usage, Advertising and Data Privacy.”
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 10FACEBOOK'S BUSINESS SOLID DESPITE FAKE NEWS, CAMBRIDGE ANALYTICA SCANDAL
FACEBOOK WORLDWIDE REVENUES
Q1 2017 – Q1 2018 +49.0%
FACEBOOK SHARE PRICES
Jan 1, 2017 – May 1, 2018 +42.5%
Facebook says it will add The New York Times reports that
3,000 workers to monitor and Cambridge Analytica improperly used
remove violative content data from millions of Facebook users
Facebook says it is using Facebook launches Brands including Tesla
artificial intelligence (AI) to flag Watch platform for and Mozilla pause their
videos that violate its policies short series Facebook activity
DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY
2017
2018
A Facebook user Facebook announces In testimony before the US Congress,
posts live video of a monetization eligibility Facebook CEO Mark Zuckerberg
murder he committed; standards aimed at apologizes for data privacy scandal
the company admits it promoting brand safety
needs “to do better”
Facebook: We can’t stop Facebook releases first
all live stream suicides Community Standards
Enforcement report
Source: Compiled by eMarketer from company reports and AdNews, Bloomberg, NPR, New York Post, The New York Times, Reuters, US SenateTWITTER When it comes to video content specifically, Twitter has
long restricted monetization to premium content under its
Amplify partner program, which it launched in 2013. The
Twitter’s challenge with policing video is less urgent
program ensures that video ads run only against vetted,
than YouTube’s or Facebook’s, because Twitter professional content from partners that include top news,
doesn’t directly monetize tweets via video ads. sports and entertainment channels, such as Bloomberg,
However, brand safety is an ongoing concern for the NBA and A&E.
Twitter given the plethora of trolls and bots on the
platform—and the potential for users to run afoul The program has been a success for Twitter. Since early
2016, video has been the its largest revenue-generating
of company guidelines for tweeting or retweeting
ad format, and as of Q4 2017, more than half of its
offensive content. business came from video ads. Video was also the
company’s fastest-growing ad format in Q1 2018, and it
In a briefing with reporters in May 2018, Twitter CEO Jack
has recently announced new video content partnerships
Dorsey said the company would double down on problem
and more ways users can incorporate video into
accounts by examining actions such as how often people
their tweets.
tweet about accounts that don’t follow them, or whether
they’ve confirmed their email address. According to Despite Twitter’s success with video, some experts say
coverage of the meeting in Reuters, tweets from suspect the company is playing it safe by restricting monetization
accounts will appear lower in some areas of the service, to professional content.
such as search results or replies, even if the tweets
themselves have not been found to violate rules. “Twitter hasn’t solved the problem. They’ve essentially
just shrunk their universe to a controlled environment,”
“We want to take the burden of the work off the people said Chris Innes, COO at ad tech firm SteelHouse. “That
receiving the abuse or the harassment,” Dorsey said, may work for a short period of time, but they have to
acknowledging that past efforts to fight abuse “felt like build a real solution. At some point they’re going to have
whack-a-mole.” to open up to every other SMB [small and medium-sized
business] advertiser.”
Part of Twitter’s effort to improve brand safety involves
combatting trolls “that distort and detract from the public Although Amplify is theoretically open to any advertiser
conversation on Twitter,” according to a May 15, 2018, that abides by Twitter’s terms of service, the program
blog post titled “Serving Healthy Conversation.” The caters mainly to national brands that are looking to attach
post indicated that fewer than 1% of Twitter accounts themselves to premium content from publishing partners
are reported for abuse, but that much of the reported including Bloomberg Television, the Discovery Channel,
behavior doesn’t violate company standards. Still, the NFL and the FIFA World Cup. Advertisers that have
accounts that get flagged for violations, whether they’ve sponsored NFL programming through Twitter Amplify
committed them or not, can have a disproportionate and include McDonald’s, Verizon and Microsoft.
negative impact on people’s experience on Twitter.
Like YouTube and Facebook, Twitter has posted strong
The company said it would start considering a wide results since early 2017, indicating that its brand safety
variety of signals to deprioritize potentially problematic challenges and relatively conservative approach to video
accounts before they get reported by other users. monetization have not hurt its bottom line or its valuation.
In Q1 2018, Twitter posted earnings of $665 million, or
“In our early testing in markets around the world, we’ve
21.3% better than its Q1 2017 figure of $548 million. In
already seen this new approach have a positive impact,
that time fame, its share price has risen by 93.2%,
resulting in a 4% drop in abuse reports from search
closing at $34.04 on May 1, 2018, up from $17.62 on
and 8% fewer abuse reports from conversations,” the
January 1, 2017.
company wrote in the blog post. “That means fewer
people are seeing tweets that disrupt their experience
on Twitter.”
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 12TWITTER'S BOTTOM LINE UNHURT BY BRAND SAFETY CHALLENGES
TWITTER WORLDWIDE REVENUES
Q1 2017 – Q1 2018 +21.3%
TWITTER SHARE PRICES
Jan 1, 2017 – May 1, 2018 +93.2% Twitter unveils new
Twitter updates election security
its rules governing measures ahead of
prohibited content 2018 midterms
Twitter starts Twitter suspends account of President Trump
testing pre-roll neo-Nazi site Daily Stormer retweets incendiary A federal judge
video ads with after white supremacist rally in anti-Muslim videos rules Trump's
Dentsu, stressing Charlottesville, Va. from controversial blocking of Twitter
its brand-safe group users who disagree
premium video with him is
content unconstitutional
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY
2017
2018
A couple in Finland Twitter issues Twitter launches Twitter says more
accidentally live streams their update of its an alpha test of a than half its business
death on Twitter's Periscope efforts to combat programmatic came from video in
video platform misinformation in advertising and Q4 2017
the wake of the real-time bidding
2016 presidential (RTB) platform
election
In congressional testimony,
Twitter says it is “determined to meet”
the challenge of abuse on its platform
Source: Compiled by eMarketer from company reports and AdExchanger, Business Insider, The Daily Mail, NPR, The New York Times, Recode, The HillBRAND SAFETY: A An April 2018 Netflix statement responding to the most
recent case of YouTube inadvertently placing ads next
SHARED RESPONSIBILITY to violative content alluded to this shared responsibility:
“We employ numerous filters to avoid having our content
As YouTube and social media companies grapple with appear on sites or videos that clearly don’t represent us or
buttressing their platforms against abuse and other our values,” read the statement. “While that works well
forms of problematic content, a marketer-side study most of the time, there are a small number of instances
illustrates the scope of the challenge from a different where it doesn’t and we are working closely with Google
to close that gap further.”
vantage point.
The statement refers to a joint effort by Netflix and
A November 2017 GumGum survey conducted by
Google and implicitly acknowledges that perfection is
Custom indicated that two-thirds of US marketers have
not a realistic goal. Rather, the companies are working
had their brands, or their clients’ brands, exposed to brand
to “close the gap further,” a tacit recognition that the
safety issues. For 43% of respondents, these issues have
best anyone can hope is to minimize—not eliminate—
occurred a few times, while 9% have experienced them
brand risk.
regularly and 15% once. Only 33% of participants said
they’d never had these issues.
Andrea Ching, CMO at OpenSlate, also drove home the
point about brands’ role in ensuring safety.
Frequency with Which Their Brand* Has Been
Exposed to Brand Safety Issues According to US “It’s critically important that brands lean into
Marketers, Nov 2017
% of respondents
understanding their own suitability guidelines and make
sure they have a strategy in place that where they
Regularly advertise meets their standards,” she said. “Brands and
9% marketers have to make sure they have the types of
Never
33% partners that can help them deliver on those standards
across all these major platforms at scale.”
A few times
43%
Once
15%
KEY TAKEAWAYS FOR MARKETERS
Note: includes agencies, brands and digital publishers; *or their clients ■■ The risk to brands of having their content placed
Source: GumGum, "The New Brand Safety Crisis" conducted by Custom (a
Digiday Media Agency), Jan 23, 2018 next to offensive content is real and should not be
235013 www.eMarketer.com minimized. Brands live and die by their reputations,
While it’s anyone’s guess whether the level of scrutiny so they are right to jealously guard their public image
that’s been placed on video and social media platforms across traditional, digital and social media.
will eventually result in more US government regulation, ■■ YouTube, Facebook and Twitter have all stepped
some industry executives anticipate a tighter regulatory up their policing efforts, and the communication
framework, which could lead to more stability and trust in strategies around them. These platforms will continue
the platforms. to hire more staff as new situations arise and, more
to the point, refine their machine learning algorithms
Michael Priem, CEO of Modern Impact, said: “We
to better catch violative videos and other pieces of
inevitably are going to see a bunch of new regulations,
content before they are seen. Nevertheless, none will
and it’s been an ongoing idea. As we’ve evolved and
ever be 100% safe while they continue to host user-
become more sophisticated, we need to have guidelines.
generated media.
I don’t think that regulations, in nature, always mean
they’re going to be bad or negative.” ■■ Brand safety issues have had little or no apparent
effect on top-line revenues or valuations at the three
And even though most of the media coverage around companies in question. Their business results have not
brand safety and other issues has focused on the seemed to suffer as a result of the many challenges
platforms themselves, marketers also play a part in they’ve faced since early 2017.
ensuring that brands are kept safe in the digital jungle.
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 14■■ Marketers continue to see value in these platforms, Matthew Fanelli
which are the core of most digital video advertising Senior Vice President
strategies. For better or worse, the duopoly of Google MNI Targeted Media
and Facebook continues to thrive. Still, marketers need Interview conducted on May 21, 2018
to continue to monitor usage and consumer attitudes
toward these platforms, and be prepared to react Christopher Hansen
quickly to any changes. Chief Product Officer
IgnitionOne
■■ Although the onus is on media platforms to provide Interview conducted on May 9, 2018
brand-safe environments, advertisers also have a
role to play in monitoring where and how their brand Mike Henry
messages appear. CEO
OpenSlate
■■ While the business impact of brand safety issues
Interview conducted on May 10, 2018
appears to be limited so far, caution is in order for
marketers. New issues have cropped up in Q2 2018
Chris Innes
that could affect companies’ bottom lines and share
COO
prices in upcoming earnings reports, and there’s no
SteelHouse
telling what could happen if these issues accelerate
Interview conducted on May 4, 2018
or new ones arise. And even if there’s no short-term
impact on business results, the cumulative effect of
Todd Krizelman
brand safety challenges could erode trust in platforms
CEO
and drive down ad spending.
MediaRadar
Interview conducted on May 16, 2018
EMARKETER INTERVIEWS Dino Myers-Lamptey
UK Managing Director
The Next Battle for Brand Safety Is in the MullenLowe Mediahub
Comment Section Interview conducted on May 8, 2018
Susan Schiekofer
Michael Priem
Chief Digital Investment Officer
CEO
GroupM
Modern Impact
Interview conducted on May 2, 2018
Interview conducted on May 9, 2018
Are Brands Shifting Ad Dollars After
Cambridge Analytica? Tuomo Riekki
CTO and Founder
Angela Yang Smartly.io
Director, Connections Interview conducted on May 4, 2018
T3
Interview conducted on May 8, 2018
Victor Wong
CEO
Serena Budhwani Thunder
Account Director Interview conducted on May 10, 2018
Sandbox
Interview conducted on May 7, 2018
Andrea Ching
CMO
OpenSlate
Interview conducted on May 10, 2018
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 15READ NEXT EDITORIAL AND
PRODUCTION CONTRIBUTORS
Changes to Facebook Advertising After Cambridge
Analytica: What’s Happening with Usage, Advertising Michael Balletti Copy Editor
and Data Privacy Katie Hamblin Chart Editorial Manager
Facebook Advertising 2018: eMarketer Tackles Big Dana Hill Director of Production
Questions on Usage, Privacy and Ad Targeting in Wake Ann Marie Kerwin Executive Editor, Content Strategy
of Cambridge Analytica Stephanie Meyer Senior Production Artist
Digital Video Advertising Best Practices 2018: Heather Price Managing Editor, Content
Connecting the Dots Across Screens and Platforms Erica Walker Copy Editor
Global Digital Video Viewers: eMarketer’s Estimates
and Forecast for 2016—2021, with YouTube and
Mobile Video Numbers
Q2 2018 Digital Video Trends: Monetization, Audience,
Platforms and Content
SOURCES
AdExchanger
AdNews
Bloomberg
Business Insider
CNBC
CNET
CNN
Facebook
Google
GumGum
MarketWatch
New York Post
National Public Radio (NPR)
Recode
Reuters
The Daily Mail
The Hill
The New York Times
The Wall Street Journal
Twitter
US Senate
Variety
YouTube
POLICING VIDEO CONTENT ON YOUTUBE, FACEBOOK AND TWITTER ©2018 EMARKETER INC. ALL RIGHTS RESERVED 16The leading research firm for marketing in a digital world.
Coverage of a Digital World
eMarketer data and insights address how consumers
spend time and money, and what marketers are doing to
reach them in today’s digital world. Get a deeper look at
eMarketer coverage, including our reports, benchmarks and
forecasts, and charts.
Confidence in the Numbers
Our unique approach of analyzing data from multiple
research sources provides our customers with the most
definitive answers available about the marketplace.
Learn why.
Customer Stories
The world’s top companies across every industry look to
eMarketer first for information on digital marketing, media
and commerce. Read more about how our clients use
eMarketer to make smarter decisions.
Your account team is here to help:
Email research_requests@emarketer.com to submit a request for research support, or contact
accounts@emarketer.com or 866-345-3864 to discuss any details related to your account.You can also read
